CHOOSEOther
Jev in Texas Hold'em
Ariel Shadkhan@ArielSh
2:59The author tested Jev's ability to assist in playing Texas Hold'em, finding it did not outperform a basic algorithm and was not worth the extra cost.
A moment before Kippur, I was looking for one last sin. Jumping on the current hype train with Jev, I decided to take it for a ride.
And there is no better sin than... cheating 😈
So I put it to the test, taking Jev as my assistant for a good old Texas Hold’em game :]
The result... 🥁 Well, it was able to help me win a tournament (at the last moment, taking me from underdog to taking over the tournament).
But... was it any better than a simple algorithm that plays strong hands, folds when there’s nothing there, and backs off under pressure? Well... nope.
So the extra token cost (even though it’s super cheap) just wasn’t worth it.
I guess that for my next sin, I would need to take Jev to something where it can actually beat basic decision-making (if/else) algorithms... cheating bots in games were here way before LLMs or “System One” models.
Maybe I just didn’t use it correctly? Maybe it’s better at analyzing player behavior in real time? Don’t know. Worth trying...
